pyg-team
Request to add two graph self-supervised methods
🚀 Feature
Request to add two graph self-supervised methods to Pytorch Geometric: (1) heuristic designed graph self-supervised tasks (ICML'20): https://arxiv.org/abs/2006.09136; code: https://github.com/Shen-Lab/SS-GCNs; (2) contrastive learning as self-supervision (NeurIPS'20): https://arxiv.org/abs/2010.13902; code: https://github.com/Shen-Lab/GraphCL.
Motivation
Self-supervision as an emerging technique has been employed to train graph neural networks (GNNs) for more transferrable, generalizable, and robust representation learning of graphs.
Additional context
@rusty1s Hello, I implemented the GraphCL model with sparse_tesnor support. I use the PyGCL code, and rewrite some code. Currently, I think PyG can not support the contrastive learning model. How should I contribute it to PyG?
That sounds cool :) We are very happy to take a pull request in on that one, including both the model and an example. Let me know if you have any questions.
@kou18n I'd like to work on this again. Can I take it up?
Of course. I haven't submitted the code. Thank you for your help. Maybe we can finish it together.
@kou18n I'd like to work on this again. Can I take it up?
Of course. I haven't submitted the code. Thank you for your help. Maybe we can finish it together.
Sounds good. I'll create a new PR and we can discuss async.